1. Field of the Invention
The present invention relates to a method of producing a semiconductor device, in particular, a power semiconductor device that can be applied to power conversion apparatuses, has a thickness between 90 to 200 μm, and is provided with an electrode on the rear surface of the device.
2. Description of Related Art
An IGBT (insulated gate bipolar transistor), a type of power semiconductor device, is a one-chip power device that gives a high speed switching characteristic and voltage-driving ability, which are specialties of MOSFETs (or insulated gate type field effect transistors), and a low forward voltage drop characteristic, which is a specialty of bipolar transistors. Applications of IGBTs have been expanded from industrial fields such as general purpose inverters, AC servo systems, uninterruptible power supplies (UPS), and switching power supplies, to consumer appliances such as microwave ovens, electric rice cookers, and electric flash lights. IGBTs with low forward voltage drop are being developed employing newly devised chip structures and promoting loss reduction and efficiency improvement of apparatuses using the IGBTs.
IGBTs include punch-through type (PT type), non punch-through type (NPT type), and field stop type (FS type). A common type of structure is an n channel, vertical double diffusion structure, except in special applications. Accordingly, the following description will be based on examples of n channel IGBTs, although the description is similarly applicable to p channel IGBTs.
A PT type IGBT uses an epitaxial substrate that is formed by epitaxially growing an n+ buffer layer and an n− active layer on a p+ semiconductor substrate. In an example of a device having a withstand voltage of 600 V, while a thickness of about 100 μm is enough for an active layer, an overall thickness including the p+ semiconductor substrate needs to be a relatively large value in the range of 200 to 300 μm. Moreover, the epitaxial substrate is costly.
For reducing costs, NPT type or FS type IGBTs using an FZ substrate instead of the epitaxial substrate have been developed. The FZ substrate is cut from a semiconductor ingot made by a floating zone (FZ) process. These types of IGBTs have a shallow p+ collector layer with low dose (low injection p+ collector) formed in the rear surface region of the device.
FIG. 1 is a sectional view showing a structure of an NPT type IGBT produced using an FZ substrate. As shown in FIG. 1, an n− semiconductor substrate composed of an FZ substrate is used for an active layer 1 and both a p+ base region 2 and an n+ emitter region 3 are selectively formed in the front surface region of the substrate. On the substrate surface, a gate electrode 5 is formed through a gate oxide film 4. An emitter electrode 6 is in contact with the emitter region 3 and the base region 2, and insulated from the gate electrode 5 by the interlayer insulation film 7.
A p+ collector layer 8 and a collector electrode 9 are formed in the rear surface region of the substrate. The overall thickness of the substrate is significantly thinner in an NPT type than in a PT type. Hole injection rate being controllable, high speed switching is possible without life time control. Use of an FZ substrate in place of an epitaxial substrate achieves a low cost.
FIG. 2 is a sectional view showing a structure of an FS type IGBT. As shown in FIG. 2, the device structure in the front surface region of the substrate is the same as that in the NPT type of FIG. 1. In the rear surface region of the substrate, an n buffer layer 10 is provided between an n− active layer 1 and a p+ collector layer 8. An overall thickness of the substrate in the range of 100 μm to 200 μm is achieved in the FS type by using an FZ substrate.
The active layer 1, being depleted similarly to the case of the PT type, has a thickness of about 100 μm in a device with a withstand voltage of 600 V. Life time control is not needed as in the case of the NPT type. In order to further reduce a forward voltage drop, a type of IGBT combining a trench structure and an FS type structure has been recently proposed. The trench structure has a narrow, deep groove formed in the chip surface region and a MOSFET structure beside the groove.
In producing an FS type IGBT using an FZ substrate, a surface device structure is first fabricated in the substrate surface region. After that, the rear surface of the substrate is ground to make the substrate thinner. Then, two types of ions are injected from the rear surface of the substrate with a reduced thickness, to form a buffer layer 10 and a collector layer 8 in the rear surface region of the substrate through an activation heat treatment. Finally, a collector electrode 9 is formed of aluminum or another metal on the surface of collector layer 8 by evaporation or sputtering.
There is also a need for IGBTs exhibiting reverse withstand ability (reverse blocking IGBTs), suitable for use in matrix converters. An n channel reverse blocking IGBT, for example, has a high concentration p type isolation region formed in the side region of a normal n channel IGBT and connecting to a collector layer. In producing a reverse blocking IGBT using an FZ substrate, an isolation region is first formed by selective diffusion of impurities from the front surface of the substrate. After that, similar to the case of the FS type IGBT, sequentially conducted are: fabrication of a surface device structure, grinding the rear surface of the substrate, ion implantation into the rear surface region of the substrate, activation heat treatment, and evaporation or sputtering to form a collector electrode.
In an FS type IGBT, the buffer layer is subjected to a high electric field when a forward bias voltage is applied; in a reverse blocking IGBT, a PN junction at the rear surface side of the device is subjected to a high electric field when a reverse bias voltage is applied. Because the PN junction in these devices is located at a shallow depth of about 0.3 μm in the rear surface region, a small flaw in the rear surface region can cause a punch-through phenomenon leading to loss of device function.
Evaporation or sputtering of a metal such as aluminum for the collector electrode is apt to generate spikes 11 of the metal protruding to the silicon substrate at the interface between a collector layer 8 of silicon and a collector electrode 9 of a metallic electrode as shown in FIG. 3. If a spike 11 reaches a buffer layer 10 in an FS type IGBT, unfavorable leakage current results. If a spike 11 reaches a PN junction in a reverse blocking type IGBT, insufficient reverse withstand voltage and unfavorable reverse leakage current result. It is essential for the above-mentioned thin IGBT having a thin collector layer to be formed with a metallic electrode on the rear surface of the substrate to avoid generation of the spikes 11 in order to reduce the proportion of defective devices.
Japanese Patent Unexamined Publication No. 2002-343980 discloses a discrete variable capacitance diode for use in high frequency circuits in which spikes of aluminum into a silicon diffusion layer are avoided by forming an anode electrode using aluminum-silicon with a silicon concentration of 3% to 5%. Japanese Patent Unexamined Publication No. 2002-343980 also discloses that an aluminum-silicon electrode with a silicon concentration in the range of 1 to 2% has been used to avoid aluminum spikes in large scale integrated circuits (LSI) according to the prior art.
The aluminum-silicon electrode disclosed in Japanese Patent Unexamined Publication No. 2002-343980, however, is especially suited for discrete variable capacitance diodes or LSIs, but not suited to avoiding generation of aluminum spikes in a rear surface electrode of power semiconductor devices such as IGBTs. In particular, in a power semiconductor device in which ion implantation has been conducted on a rear surface of the substrate made thin by a back grinding process and a shallow impurity diffusion layer has been formed along the rear surface of the substrate, the silicon concentration in the aluminum-silicon electrode must be optimized to avoid generation of aluminum spikes. Further, because the substrate after the back grinding is very thin, a thick aluminum-silicon electrode causes warping in the substrate and undesirably generates cracks in the substrate. Therefore, the thickness of the aluminum-silicon electrode must be optimized, too.